Hi Ming, to clarify my comments to your patch to introduce dead_hctx_list here's a small patchset to address my concerns. If you agree it would possibly be best to fold it into your patchset. Patches are relative to the 'nvme-5.2' branch of the nvme git repository plus the 'V6' iteration of Mings patchset 'blk-mq: fix races related with freeing queue'. Hannes Reinecke (5): blk-mq: rename 'dead_hctx_XX' to 'unused_hctx_XX' blk-mq: drop 'hctx_idx' argument from blk_mq_alloc_hctx() blk-mq: remove 'nr_queue' argument in blk_mq_exit_hw_queues() blk-mq: Set hctx pointer to NULL in blk_mq_exit_hw_queues() blk-mq: free q->queue_hw_ctx in blk_mq_exit_queue() block/blk-mq.c | 45 +++++++++++++++++++++++++-------------------- block/blk-mq.h | 2 +- include/linux/blkdev.h | 6 +++--- 3 files changed, 29 insertions(+), 24 deletions(-) -- 2.16.4
